Integrating machine learning with lateral flow immunoassay for ultrafast and sensitive tadalafil detection

Tadalafil, a phosphodiesterase type 5 inhibitor frequently detected in functional foods and dietary supplements, poses significant risks. To enable sensitive on-site detection, an anti-tadalafil monoclonal antibody (3C5) with an IC₅₀ of 0.32 ng mL^-1 was prepared. Antibody-antigen binding kinetics revealed distinct variations in the early detection stage, prompting the integration of lateral flow immunoassay with machine learning (LFIA@ML) to further speed up testing. A Random Forest model trained on multi-dimensional signal features achieved fast detection and high accuracy, with a coefficient of determination of 0.995 and a mean absolute error of 0.038. For 140 samples, the LFIA@ML demonstrated a 94.3 % prediction accuracy and significantly reduced detection time to 3 min. The assay exhibited a strong correlation with LC-MS/MS, with recoveries ranging from 92.3 % to 112 % (coefficient variations ≤15.37 %). The LFIA@ML showcased the ability to reduce detection time by using the Random Forest algorithm and was validated in wine samples.